Monetary Policy in Emerging Market Economies: What Lessons From the Global Financial Crisis?

Board of Governors of the Federal Reserve System International Finance Discussion Papers Number 1042 February 2012 Monetary Policy in Emerging Market Economies: What Lessons from the Global Financial Crisis? Brahima Coulibaly* NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References to International Finance Discussion Papers (other than an acknowledgment that the writer has had access to unpublished material) should be cleared with the author or authors. Recent IFDPs are available on the Web at www.federalreserve.gov/pubs/ifdp/. This paper can be downloaded without charge from Social Science Research Network electronic library at www.ssrn.com.

Monetary Policy in Emerging Market Economies: What Lessons from the Global Financial Crisis? Brahima Coulibaly* February 2012 Abstract During the 2008-2009 global financial crisis, emerging market economies (EMEs) loosened monetary policy considerably to cushion the shock. In previous crises episodes, by contrast, EMEs generally had to tighten monetary policy to defend the value of their currencies, to contain capital flight, and to bolster policy credibility. Our study aims to understand the factors that enabled this remarkable shift in monetary policy, and also to assess whether this marks a new era in which EMEs can now conduct countercyclical policy, more in line with advanced economies. The results indicate statistically significant linkages between some characteristics of the economies and their ability to conduct countercyclical monetary policy. We find that macroeconomic fundamentals and lower vulnerabilities, openness to trade, and international capital flows, financial reforms, and the adoption of inflation targeting all facilitated the conduct of countercyclical policy. Of these factors, the most important have been the financial reforms achieved over the past decades and the adoption of inflation targeting. As long as EMEs maintain these strong economic fundamentals, continue to reform their financial sector, and adopt credible and transparent monetary policy frameworks such as inflation targeting, the conduct of countercyclical monetary policy will likely be sustainable. Keywords: Monetary policy, crises, macroeconomic stabilization JEL classifications: E52, E58, E63 *Senior Economist in the Division of International Finance of the Federal Reserve System. Mailing address: Division of International Finance, Board of Governors, Federal Reserve System, Mail Stop 24, Washington, DC 20551,USA; email: brahima.coulibaly@frb.gov; tel.: (202) 452-2609; fax: (202) 736- 5638. The author thanks Shaghil Ahmed, John Rogers, Roberto Chang, and participants of the 2nd BIS CCA Conference on "Monetary Policy, Financial Stability and the Business Cycle" for helpful comments and discussions, and Andrew Brooks and Karan Jain for outstanding research assistantship. The views in this paper are solely the responsibility of the author(s) and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or of any other person associated with the Federal Reserve System.

1 Introduction During the (cid:133)nancial crisis of 2008-2009, emerging market economies (EMEs) loosened monetary policy considerably to cushion against the global (cid:133)nancial shock and to foster economic recovery. This is a remarkable departure from previous crisis episodes during which EMEs generally had to raise interest rates in order to bolster the credibility of monetary policy, to defend the value of their currencies, and to contain capital (cid:135)ight. Our study asks what factors enabled this shift in monetary policy of EMEs, and it assesses whether this shift marks a new era in which EMEs can pursue countercyclical monetary policy like their counterparts in advanced economies. Macroeconomicpolicies(cid:150)both(cid:133)scalandmonetary(cid:150)tendtobecountercyclicalinadvancedeconomies. In EMEs, by contrast, these policies tend to be procyclical or, at best, acyclical that tend to be countercyclical. This feature of monetary and (cid:133)scal policy deprived EMEs of important macroeconomic stabilization tools, and might partly explain the higher volatility of output in EMEs compared with the advanced economies documented in Aguiar and Gopinath (2007) and others. One way to reduce output volatility and enhance welfare in EMEs is to understand the factors that prevented policymakers in EMEs from conducting countercyclical policy in the past, and to devise policies to help them use (cid:133)scal and monetary policy for macroeconomic stabilization. Studies have analyzed the factors driving the cyclicality in the (cid:133)scal policy of EMEs. See for example,GavinandPerotti(1997),TalviandVegh (2004),andothers.1 Bycontrast,therearefew empiricalstudiesofthecyclicalityinthemonetarypolicyinEMEs. Thissparsitylikelyre(cid:135)ectsthe di¢ culty of (cid:133)nding common monetary policy instruments over time and across countries, as these instruments depend importantly on the exchange rate regime. Even with common instruments, characterizing the monetary policy stance is di¢ cult. In Kaminsky, Reinhart, and Vegh (2004) examines cyclicality in the monetary policy of a broad set of countries covering both emerging market and advanced economies by relying primarily on short-term interest rates. Assuming imperfect substitution between domestic and foreign assets, short-term interest rates can represent common monetary policy instruments under both (cid:135)exible and predetermined exchange rate regimes. Using these short-term interest rates, Kaminsky et al. (2004) estimate a Taylor rule policy function for each country and (cid:133)nd that monetary policy is generally countercyclical in advanced economies. By contrast, it tends to be procyclical in EMEs. 1 Including, Braun (2001), Lane (2003), Gupta et al. (2004), Riascos and Vegh (2003), Kaminsky, Reinhart, and Vegh (2004), etc. 1

Kaminsky, Reinhart, and Vegh (2004) did not explore the factors preventing EMEs from conducting countercyclical monetary policy. These factors were explored in Calderon et al. (2003) for a set of eleven EMEs. They (cid:133)nd that the ability of these EMEs to conduct both countercyclical (cid:133)scal and monetary policies is determined by the credibility of their policies. Our study adds to this sparse literature by examining the behavior of monetary policy during economic crises. Economic crises are costly in output and welfare losses. Optimal response to crises generally require countercyclical policies to cushion the shock and to foster economic recovery. Yet in EMEs, the crises are exacerbated by procyclical policies, including monetary policy. However, during the 2008-2009 global (cid:133)nancial crisis, central banks in EMEs were able to loosen monetary policy considerably, perhaps signaling that monetary policy has evolved in these countries. To our knowledge, this is the (cid:133)rst study to comprehensively assess the factors that determined the cyclicality of monetary policy during the 2008-2009 (cid:133)nancial crisis, and during crises more generally. Weconstructalargedatasetfor188advancedandemergingmarketcountriesfrom1970through 2009. We identify 1,462 (cid:133)nancial and economic crisis years, and examine the behavior of monetary policy during those crises. The results con(cid:133)rm that advanced economies have historically conducted countercyclical monetary policy during crises while EMEs tended to tighten monetary policy. However,thedi⁄erenceinpolicyresponsebetweenthetwosetsofcountrieshasbeenfading. In the most recent decade, notably during the 2008-2009 crisis, EMEs have generally conducted countercyclical policy like their counterparts in the advanced economies. Our estimation strategy uses a Logit regression model to examine the factors that have facilitated the conduct of countercyclical monetary policy in EMEs. The results indicate statistically signi(cid:133)cant linkages between some characteristics of the economies and policymakers(cid:146)ability to conduct countercyclical monetary policy. We (cid:133)nd that while stronger macroeconomic fundamentals, reduced vulnerabilities, greater openness to trade and international capital (cid:135)ows facilitated the conduct of countercyclical policy, the most important determinants have been the (cid:133)nancial reforms achieved over the past decades and the adoption of in(cid:135)ation targeting. In(cid:135)ation targeting regimes, whicharebecomingmorepervasiveamongEMEs, enhancegreaterpolicytransparencyand(cid:135)exibilityofmonetarypolicy. EMEsalsoachievegreaterpolicycredibilitybyadoptingin(cid:135)ationtargeting regimes and by achieving greater (cid:133)nancial reforms. As long as EMEs maintain strong economic fundamentals, continue to reform their (cid:133)nancial markets, and adopt credible and transparent mon- 2

etary policy frameworks such as in(cid:135)ation targeting, the conduct of countercyclical policy as an economic stabilization tool might be sustainable. The remainder of the paper is organized as follows: In the next section, we discuss some of the literature on the determinants of monetary policy stance in EMEs. Sections 4 and 5 describe the econometric strategy, the data, and the results. Section 5 is devoted to caveats and robustness analyses, and we o⁄er concluding remarks in Section 6. 2 Determinants of Monetary Policy in Emerging Market Economies During Crises During economic crises the common policy prescription is to loosen monetary policy in order to support domestic economic activity. This prescription is theoretically motivated by the Keynesian models and illustrated in practice by the Taylor rule type of approach to monetary policy. In this setting, looser monetary policy is necessary to help close the negative output gap and restore full employment. The consequent increase in domestic liquidity tempers the e⁄ect of the contraction inexternalcreditthatusuallyoccursduringEMEs(cid:146)crises. Advancedeconomieshavegenerallyfollowedthispractice. InEMEs, however, otherfactorshavepreventedtheconductofcountercyclical policy or made countercyclical policy undesirable. Conditional on speci(cid:133)c economic vulnerabilities, countercyclical policy might not be optimal. For example, if a country has a large fraction of its debt that is short term and denominated in foreign currency, the adverse balance sheet e⁄ects of an exchange rate depreciation induced by a countercyclical policy could more than o⁄set any potential costs of a procyclical policy. In this case, it would be optimal to maintain a procyclical monetary policy. Internal vulnerabilities such as these or other institutional de(cid:133)ciencies explain the inability or undesirability of policymakers in EMEs to conduct countercyclical policy. And authorities in these countries have often been more concerned about bolstering the credibility of policy, containing capital (cid:135)ight, and defending the values of their currencies. AstudybyCalderonetal.(2003)ofthecyclicalityofmonetarypolicyinsomeEMEs, (cid:133)ndsthat credibilityofpolicywasthedeterminingfactor. AspointedoutbyLane(2003),whenthemonetary authority lacks credibility, a temporary loosening of monetary policy is perceived as heralding a persistent switch to a loose money regime with adverse e⁄ects on con(cid:133)dence and increases in risk 3

premiums demanded by foreign investors.2 We include in our study some variables that capture the strength of institutions and the credibilityofpolicy: theexchangerateregime, anindicatorforin(cid:135)ationtargeting, ameasureof(cid:133)nancial reforms(cid:150)the extent to which authorities have allowed market forces to determine outcomes in credit and (cid:133)nancial markets(cid:150), and a measure of (cid:133)nancial development. In addition to measuring strength of institutions and the credibility of policy, the (cid:133)nancial development variable has a unique relevance. Financial development enables a more e¢ cient transmission of monetary policy and, hence, increases the incentives to conduct countercyclical policy. Also, the development of (cid:133)nancial markets has traditionally promoted more borrowing on domestic markets and in local currencies. A higher share of local currency debt reduces risks of capital (cid:135)ight, and risks of currency and maturity mismatches. As such, development of domestic (cid:133)nancial markets facilitates the conduct of countercyclical policy. Devereux and Lane (2003) (cid:133)nds that countries with a greater dependence on foreign currency debt are more likely to tailor policy to minimize exchange rate volatility with the creditor country. Besides restricting monetary policy, dependence on external debt and debt with shorter maturities has a⁄ected the perceived solvency of EMEs during crises. With this consideration in mind, we includevariablesonexternaldebtanditsmaturitystructureandvariablesonthecountry(cid:146)s(cid:133)nances such as foreign exchange reserves and central government debt. We also consider other macroeconomic fundamentals such as current account balances and in(cid:135)ation. A low in(cid:135)ation environment facilitates the loosening of monetary policy, consistent with the prescriptions from a Taylor rule function. In(cid:135)ation could also capture the independence of the central bank and, hence, credibility of monetary policy. Several studies document that central banks in lower-in(cid:135)ation countries are more independent (see for example, Alesina and Summers, 1993); and central bank independence improves the e¢ ciency of monetary policy (Mishkin, 2010). Economic integration is also an important factor. A study by Yakhin (2008) (cid:133)nds that under (cid:133)nancial integration, the optimal monetary policy is countercyclical, but procyclical under autarky. These results suggest an important role for openness. We include a measure of (cid:133)nancial openness, and also trade openness. In sum, the variables we explore in this study can be classi(cid:133)ed into four categories: Macroeco- 2See also Caballero (2002), Calvo and Reinhart (2002) and Mendoza (2002). 4

nomic fundamentals and vulnerabilities, openness, monetary policy and exchange rate framework, and (cid:133)nancial development and reforms. These variables are not independent of each other and the categories are likely not insular. In the empirical analysis, we assess the statistical link between monetary policy and these variables in both univariate and multivariate econometric frameworks. 3 Monetary Policy in Emerging Market Economies during the 2008-2009 Financial Crisis In this section, we analyze monetary policy during the 2008-2009 global crisis. At the height of the crisis, between the third quarter of 2008 and the end of the (cid:133)rst quarter of 2009, over 80 percent of EMEs loosened monetary policy. In the analysis that follows, we assess the factors that enabled most, but not all, countries to loosen monetary policy. 3.1 Econometric Speci(cid:133)cation and Data Description We estimate the following Logit model using the indicator variable for countercyclical monetary policy during crises as the dependent variable: Prob(CCMP i = 1) = (cid:8) (cid:12)0X i2007 (1) (cid:16) (cid:17) A country is considered to have conducted countercyclical monetary policy during the crisis if the cumulative change in the monetary policy rate between the third quarter of 2008 and the end of the (cid:133)rst quarter in 2009 is negative. CCMP is the indicator variable for whether country i has, i on net, lowered its monetary policy rate between the third quarter of 2008 and the end of the (cid:133)rst quarter in 2009. X represents the set of variables that determine the conduct of monetary policy. They are measured in in 2007(cid:150)the year prior to the crisis. Macroeconomic Fundamentals and Vulnerability: FXR2GDP and CAB2GDP are the foreign exchange reserves and current account balance as percent of GDP, respectively. CGD2GDP is the central government debt as percent of GDP. INF is the annual change of the consumer price index. STDT2EXTDT and STDT2FXR represent short-term external debt as percent of total external debt and foreign exchange reserves, respectively. Openness: OPENTRADE andOPENFIN capturethedegreeoftradeand(cid:133)nancialopenness, 5

respectively. TradeopennessisthethesumofimportsandexportsaspercentofGDP. For(cid:133)nancial openness, we use the Chinn-Ito index of capital account openness. It was initially introduced in Chinn and Ito (2006) and subsequently updated by the authors through 2008. The index is based on the tabulation of binary dummy variables that capture restrictions on cross-border (cid:133)nancial transactions as reported in the IMF(cid:146)s Annual Report on Exchange Arrangements and Exchange Restrictions(AREAER). Itvariesfrom-1.8to2.5,withhighernumbersindicatinggreater(cid:133)nancial openness.3 Exchange Rate Regime and Policy Credibility: IT is an indicator variable for whether the country(cid:146)s central bank is an in(cid:135)ation targeter in a given year. EXCHREG captures the rigidity of the exchange rate regime based on the classi(cid:133)cation in the IMF(cid:146)s AREAER. For a given year, each country is assigned a number between 1 and 5, with higher numbers indicating greater (cid:135)exibility of the exchange rate regime. Financial Development and Financial Reforms: FINDEV measures (cid:133)nancial development basedondataforbankdeposit,(cid:133)nancialsystemdeposits,depositbanks(cid:146)assets,andcreditextended by banks and total credit to the domestic economy. We then divide these variables by GDP and conduct a principal component analysis to obtain a single index. The resulting factor (index) is highly correlated (0.92 or higher) with the variables, and explains 93 percent of the variability of these variables. FINREF measures (cid:133)nancial reforms. It is an index constructed by Abiad et al. (2008) based on factors such as the extent of directed credit, level of reserve requirements, prevalenceofcreditcontrolsandcreditceilings,interestratecontrols,entrybarriers,capitalaccount restrictions, state ownership in banking sector, and prudential regulations and supervision of the banking sector. The index provides a number ranging from 0 to 21, with higher values indicating greater degrees of (cid:133)nancial reform. See Abiad et al. (2008) for details. In this study, we use the normalized (between 0 and 1) version of the index, also provided by the authors. The reforms indexhasahighautocorrelationcoe¢ cientof0:99. WeuseanAR(1)processtoimputethemissing values for 2006 and 2007.4 Table 1 presents summary statistics for the variables described above for the the 2008-2009 subsample. Figures 1 through 3 plot the median or mean values of some key variables through 2007 for the set of countries that lowered their policy rates (the bold line) and the set of those that did 3See http://web.pdx.edu/~ito/Readme_kaopen2008.pdf for details. 4More speci(cid:133)cally, our imputation model was FINREF =0:99 FINREF +" . it it 1 it (cid:3) (cid:0) 6

not (the thin line). Several of these key variables highlight the di⁄erence between the two sets of countries. Those that loosened monetary policy had better macroeconomic fundamentals and lower vulnerabilities: in(cid:135)ation was lower on the eve of the crisis; current account balances were in large surplus while those of countries that could not lower rates were in de(cid:133)cit; reserves as percent of GDP were higher, and central government debt as percent of GDP was lower. Also, compared with countries that could not lower rates, those that did had external debt with slightly longer maturities, and lower short-term debt as percent of reserves. They were also more open to trade and international capital (cid:135)ows, had relatively more (cid:135)exible exchange rate regimes, and were more likelytobein(cid:135)ationtargeters. Finally,countriesthatloweredmonetarypolicyhadmoredeveloped (cid:133)nancial markets and had made more progress on (cid:133)nancial reforms. Tofacilitateinterpretationofourregressionresults, wetransformallofthecontinuousvariables into categorical variables: top quartile, midquartiles, and bottom quartile. This transformation also allows for the exploration of nonlineraties and to control for the e⁄ect of potential outliers. 3.2 Estimation and Results 3.2.1 Univariate Analysis Theregressionresultsfortheunivariateanalysis(eachregressionhasonlyoneexplanatoryvariable) are shown in Table 2. Columns 2, 3, and 4 show the logit coe¢ cients, p-values, and odds ratios, respectively. Macroeconomic Fundamentals and Vulnerability: These results suggest that strong macroeconomic fundamentals and reduced vulnerability in the pre-crisis year increased the chances of conducting countercyclical monetary policy. A country with pre-crisis in(cid:135)ation in the top quartile of the distribution has lower odds of reducing rates during the crisis. Similarly, countries with the lowest government debt and highest current account surplus (in the top quartiles) were, respectively, about 2.3 and 3.5 times more likely to conduct countercyclical policy. The coe¢ cients for the share of short-term external debt, foreign exchange reserves as percent of GDP, and short-term debt as percent of foreign exchange reserves have the expected sign, but they are not statistically signi(cid:133)cant. Openness: The next set of results indicates an important role for openness in a country(cid:146)s ability to conduct countercyclical monetary policy. Countries with highly open capital accounts (top quartile of the distribution) were 3 times more likely to loosen monetary policy during the 7

2008-2009 crisis. Similarly, those most open to trade on the eve of the crisis were 2.5 times more likely to loosen monetary policy. Exchange Rate Regime and Policy Credibility: The coe¢ cient for the exchange rate regime has the expected sign; countries with the most (cid:135)exible form of exchange rate regime are more likely to loosen monetary policy but the coe¢ cient is not statistically signi(cid:133)cant. The results for in(cid:135)ation targeting, which also proxies for transparency and credibility of the central bank, are very strong. A country with an in(cid:135)ation targeting regime was about 7.6 times more likely to conduct countercyclical monetary policy than a country without one. Financial Development and Financial Reforms: The result for (cid:133)nancial reform is also very strong. It suggests that a country with the highest level of (cid:133)nancial reform was 4.5 times more likelytoloosenmonetarypolicy. For(cid:133)nancialdevelopment,thepositivecoe¢ cienthastheexpected sign though it is not statistically signi(cid:133)cant. We suspect that these variables are not necessarily independent of each other. In the next analysis, we estimate the e⁄ect of these various factors in a multivariate econometric framework. 3.2.2 Multivariate Analysis Tables 3 presents the odds ratios obtained from the multivariate regression using equation (1). We estimate the model with OxMetrics, a statistical software package that explores various combinations of regressors to maximize the (cid:133)t of the model based on the Akaike Information Criterion.5 At the outset, it suggested 5 alternative models(cid:150)Columns (1) through (5). Overall, the results are consistent with those of the univariate analysis. Countries with the lowest level of government debt on the eve of the crisis were about 2.5 times more likely to loosen monetary policy. Those that were most open, particularly to capital (cid:135)ows, had greater odds of conducting countercyclical monetarypolicy. In(cid:135)ationtargetingremainsthemostimportantdeterminantofacountry(cid:146)sability to conduct countercyclical policy. The results are strong and consistently robust across various speci(cid:133)cations. In sum, the analysis provides evidence of links between EMEs(cid:146)ability to conduct countercyclical policy during the crisis and some pre-crisis characteristics of their economies, such as level of government debt, degree of openness, and most importantly, in(cid:135)ation targeting monetary policy 5The model selection process is based on the set of variables that maximize the likelihood function and applies the Akaike Information Criterion. 8

framework. This remarkable development begs the following question: Is the ability of EMEs to conduct countercyclical policy during the 2008-09 crisis ephemeral or is it a re(cid:135)ection of structural improvements that have enabled monetary policy to become a more e⁄ective macroeconomic stabilization tool? In the next set of analyses, we explore the determinants of countercyclical monetary policy in EMEs more generally by expanding the sample to the preceding four decades(cid:150)1970 through 2009. 4 Monetary Policy in Emerging Market Economies: Beyond the 2008-2009 Crisis In this section, we explore more generally the determinants of countercyclical monetary in EMEs over the past four decades. 4.1 Identi(cid:133)cation of Crises and Monetary Policy Stance Twovariablesthatarecentraltoourstudyareindicatorsforcrisesandthemonetarypolicystance. We follow Frankel and Rose (1996) and de(cid:133)ne a crisis year as one in which the bilateral U.S. dollar exchange rate depreciated at least 25 percent, with the rate of depreciation exceeding the previous year(cid:146)sdepreciationbyatleast10percentagepoints.6 Inaddition, weincludeperiodswithnegative or zero real gross domestic product (GDP) growth in order to capture episodes of economic stress that necessitate active countercyclical monetary policy, but when exchange rate movements might not be substantial. At the outset, we obtain 1,462 episodes between 1970 and 2009. Figure 4 provides a histogram for the distribution of the crises episodes over time. The year 2009 stands out as having the most crises. There were also a higher number of crises in the early 1980s and 1990s. This tabulation is consistent with well-known economic and (cid:133)nancial crises that have a⁄ected the global economy, including the sovereign debt crises of the early 1980s, the Savings and Loans crisis and the Japanese banking crisis of the 1990s. Identifying the monetary policy stance is more complicated, primarily due to the lack of a commonmonetarypolicyinstrumentacrosscountriesandtime. Inparticular,thepolicyinstrument depends on the exchange rate regime. We follow Kaminsky, Reinhart, and Vegh (2004) and use short-term interest rates as the policy instrument. Under (cid:135)exible exchange rate regimes, short- 6We also explored two alternative de(cid:133)nitions provided by Milesi-Ferretti, Gian, and Razin (2008). 9

term interest rates characterize monetary policy since changes in money supply in(cid:135)uences these rates. However, under predetermined exchange rate regimes, short-term rates are valid monetary policy instruments only if we assume imperfect substitution between domestic and foreign assets. See, for example, Flood and Jeanne (2000) or Lahiri and Vegh (2003). For the choice of short-term rates, we begin with the monetary policy rates, and supplement with the discount or interbank rates. When these series are not available, we rely on short-term Treasury bill rates, and then money market rates. In addition to short-term interest rates, we also use growth of central banks(cid:146)domestic credit to proxy for monetary policy. Under (cid:135)exible exchange rate regimes, central bank domestic credit growtha⁄ectsthemonetarybaseandshort-termrates. Underpredeterminedexchangerateregimes and perfect substitution between domestic and foreign assets, growth in central bank credit will be o⁄set by an opposite e⁄ect in foreign exchange reserves. However, if domestic and foreign assets are imperfect substitutes, an increase in central bank credit will have some e⁄ect on the monetary base and short-term interest rates. Evenwithgoodmeasuresofthemonetarypolicyinstrument,characterizingthemonetarypolicy stanceisnotobvious. Forthepurposeofthisstudy,wede(cid:133)necountercyclicalpolicyasamovement inthedirectionoflooseningmonetarypolicyduringperiodsofeconomicstress. Wede(cid:133)neabinary indicatorvariablethattakesavalueofoneif: thepolicyratedeclinesintheyearofthecrisisrelative to the previous year or when the central bank(cid:146)s domestic credit growth in the crisis year exceeds that of the previous year and the average rate of the three years prior to the crisis. When the monetary policy rate is not available, we rely on other short-term rates. Wearemindfulofthepotentialimperfectionsassociatedwiththeuseofothershort-terminterest rates to as a measure of monetary policy. Short-term rates can change independent of the true monetary policy rate. For example, risk premia tend to increase during crises, causing some shortterm rates to rise even if policy rates have been lowered. However, in periods of crises, we posit that a decline in short-term rates likely indicates lower monetary policy rates. At the outset, we obtain the policy stance for 980 crisis years, 127 for the advanced economies and 853 for EMEs. Figure 5 presents the frequency countercyclical monetary policy during crises over time and for the two sets of countries. The (cid:133)gure highlights the contrast between the advanced economies and EMEs. While the advanced economies have traditionally conducted countercyclical monetary policy during crises, it is only in the latter periods that EMEs began to do so. During crises in 10

the 1970s, EMEs lowered rates in only about 30 percent of the crises. This fraction has increased steadily, to 70 percent in the most recent decade. During the 2008-2009 global crisis, the fraction rose further, to over 80 percent. 4.2 Econometric Speci(cid:133)cation and Data Description The econometric model is a more general version of equation (1) used in the previous section. Prob(CCMP = 1) i;(cid:28) 1 = (cid:8) (cid:12)0X i;(cid:28) 1 (2) (cid:0) (cid:0) (cid:16) (cid:17) Where(cid:8) (cid:12)0X i;j = e(cid:12)0Xi;(cid:28) (cid:0) 1 ; X i;(cid:28) 1 representsasetofexplanatoryvariablesthatcapturea 1+e(cid:12)0Xi;(cid:28) 1 (cid:0) (cid:0) country(cid:146)s abi(cid:16)lity to c(cid:17)onduct countercyclical monetary policy during crises and are measured in the year before the crisis ((cid:28) 1) for each crisis country i. The set of independent variables (X) are as (cid:0) de(cid:133)ned in the previous section, but measured with a lag. Summary statistics for the independent variables over the 1970-2009 sample period are described in Table 1. 4.3 Estimation and Results 4.3.1 Univariate Analysis Table 4 presents the regression results for the univariate model. They are generally similar to those in Table 2. Macroeconomic Fundamentals and Vulnerability: Strong macroeconomic fundamentals and reduced vulnerability increase the chances of conducting countercyclical monetary policy. A country withpre-crisisin(cid:135)ationinthebottomquartileofthedistributionis62percentmorelikelytoreduce ratesduringthecrisis. Similarly, countrieswiththelargestamountofforeignexchangereserves(in the top quartile) are about 2.5 times more likely to conduct countercyclical policy. Those with the highest levels of short-term external debt to foreign exchange reserves ratio are less likely to conduct countercyclical policy during crises. The coe¢ cients on the share of short-term external debt, current account surpluses, and government debt have the expected sign but are not statistically signi(cid:133)cant. Openness: The next set of results examine the role of openness. Overall, they suggest an important role for openness in a country(cid:146)s ability to conduct countercyclical monetary policy. Countries with highly open capital accounts (top quartile of the distribution) are 45 percent more 11

likely to loosen monetary policy during crises. Similarly, those most closed to trade are about 40 percent less likely to loosen monetary policy. Exchange Rate Regime and Policy Credibility: The coe¢ cient for the exchange rate regime has the expected sign; countries with the most (cid:135)exible form of exchange rate regime are more likely to loosen monetary policy but the coe¢ cient is not statistically signi(cid:133)cant. Again, the results for in(cid:135)ation targeting, which also proxies for transparency and credibility of the central bank, are the strongest. They suggest that a country with in(cid:135)ation targeting is nearly 7 times more likely to conduct countercyclical monetary policy than a country without an in(cid:135)ation targeting regime. Financial Development and Financial Reforms: Both (cid:133)nancial development and reforms enhance the ability to conduct countercyclical monetary policy. Countries that have achieved the highest level of (cid:133)nancial reforms are more than twice as likely to loosen monetary policy, and those with the most developed (cid:133)nancial system are 50 percent more likely to loosen monetary policy. In sum, these results suggest strong linkages between a country(cid:146)s ability to conduct monetary policy and its macroeconomic fundamentals and vulnerability, its degree of openness, the exchange rate regime and the credibility of the central bank(cid:146)s policy, as well as the degree of (cid:133)nancial development and reforms. Judging by the size of the coe¢ cients, in(cid:135)ation targeting appears to be the most important determinant of the ability to conduct countercyclical monetary policy, followed by a high level of (cid:133)nancial reforms, large amounts of foreign exchange reserves, and low in(cid:135)ation. 4.3.2 Multivariate Analysis Table 5 presents the odds ratios obtained from the multivariate regression using equation (2). We estimate the model with OxMetrics. It explored 450 models (combinations of regressors), and selected, basedontheAkaikeInformationCriterion, the12comparablealternativemodelsreported in columns 1 through 12 of the table. Macroeconomic Fundamentals and Vulnerability: Asfoundpreviously, strongermacroeconomic fundamentals and low vulnerability enhance the odds of countercyclical monetary policy. Countries with the lowest pre-crisis rate of in(cid:135)ation are more than twice as likely to lower interest rates during crises. These results are consistent with the prediction from a Taylor rule reaction function. Indeed, in a low in(cid:135)ation environment, monetary authorities can loosen monetary policy to stimulate economic activity without concerns of fueling in(cid:135)ation. We (cid:133)nd evidence that higher foreign exchange reserves as a percent of GDP enhance the odds of conducting countercyclical monetary 12

policy. Havingforeignexchangereservestocovertheexternalshort-termdebtisarobustindicator ofacountry(cid:146)sabilitytoconductcountercyclicalmonetarypolicy. Countriesinthelowestquartileof the short-term debt to foreign exchange reserves distribution are roughly twice as likely to conduct countercyclical monetary policy, and the e⁄ect appears to be monotonic. The extent to which a country can cover its short-term debt is indeed an important indicator of its solvency in periods of crises when the rollover of debt or issuance of new debt becomes di¢ cult. Openness: In one of the speci(cid:133)cations, we (cid:133)nd evidence that (cid:133)nancial openness increases the likelihood of countercyclical monetary policy. Countries most open to trade are 50 percent more likelytoloosenmonetarypolicyduringacrisis. Thecoe¢ cientfortradeopennesshastheexpected sign but it is not statistically signi(cid:133)cant. Exchange Rate Regime and Policy Credibility: The coe¢ cient for the exchange rate regime is statistically insigni(cid:133)cant. By contrast, as documented previously, in(cid:135)ation targeting remains the most robust predictor of a country(cid:146)s ability to conduct countercyclical monetary policy. In(cid:135)ation targeters are about 6-to-11 times more likely than non-targeters to loosen monetary policy during a crisis, and this e⁄ect is consistently robust across the various alternative models. Financial Development and Financial Reforms: The coe¢ cient for (cid:133)nancial reforms is robust across a number of alternative speci(cid:133)cations. Countries with the highest level of (cid:133)nancial reforms areroughly3timesmorelikelytoconductcountercyclicalmonetarypolicy. Theresultsfor(cid:133)nancial development are not signi(cid:133)cant in a number of cases but, where signi(cid:133)cant, they are counterintuitive. 5 Caveats and Robustness Analysis In this section, we conduct robustness analysis to assess the importance of some of the assumptions we have made and discuss some possible caveats. Duringtheanalysis,asinKaminsky,Reinhart,andVegh(2004),weassumethatunderimperfect substitutionbetweenforeignanddomesticassets,short-terminterestratesaregoodmonetarypolicy instruments under predetermined exchange rate regimes. To assess how this assumption a⁄ects our results, we restrict the sample to non-pegged exchange rate regimes in the (cid:133)rst robustness test. The second test, we restrict the measurement of monetary policy to policy rates and discount rates only(cid:150)the two most reliable measures(cid:150)in order to control for the e⁄ect of potential imperfections in 13

other measures of monetary policy. In the third robustness test, we remove from the sample crises episodesduringwhichpolicywasacyclical(cid:150)wheninterestratesdidnotchangebetweenthepre-crisis year and crisis year. In the last robustness test, we remove from the sample the second of any two crises that occur in consecutive years for the same country to ensure that our results are not driven by a possible doublecounting of the same crisis. The results for these robustness analyses are presented in Table 6, column 1, 2, 3, and 4, respectively. Our main results, highlighting the importance of (cid:133)nancial reforms and in(cid:135)ation targeting regimes on a country(cid:146)s ability to conduct countercyclical monetary policy, still hold. Onecaveatiswhethernominalinterestrates(notrealinterestrates)aretheappropriatemeasure of monetary policy stance. We are unable to measure in(cid:135)ation expectations and formally conduct this robustness analysis. Our study is more concerned with the direction of monetary policy from the standpoint of the central bank and not with the actual policy stance. As such, the use of nominal interest rates is appropriate. Another caveat pertains to the other nonconventional monetary policy instruments that EMEs often use. In advanced economies with well-functioning (cid:133)nancial markets, the main monetary policy instrument consists of open market operations and, to a lesser extent, adjustments to the discount rate and reserve requirement ratios. In EMEs, where (cid:133)nancial markets are underdeveloped, monetary policy use other nonconventional instruments such as credit ceilings, and moral suasion. Although this study does not take into account all of the measures of monetary policy, we believe that if data were available, changes in these instruments would generally be consistent with the changes in short-term rates. For example, it is not likely that the central bank will lower short-term interest rates and at the same time raise reserve requirements or lower the credit ceilings. We further assured that our main results are robust to a number of the caveats mentioned earlier by the consistency between the analysis over the 1970-2009 sample and the 2008-2009 subsample. In the 2008-2009 sub-sample, we have better measures of the monetary policy rates and, hence, rely less on other short-term interest rates as proxies for policy rates. Moreover, fewer countries had pegged exchange rate regimes suggesting that the assumption of imperfect substitution domestic and foreign assets is not as necessary. Finally, one might be concerned about the e⁄ect of di⁄erences in nature of crises in our sample 1970-2009. Again, we are comforted by the consistency of the results obtained from the two samples. The cross-section analysis that uses 14

only the 2008-2009 sub-sample allows us to control for the nature of the crisis and identify the determinants of countercyclical policy through cross-country variations. 6 Concluding Remarks During the recent global (cid:133)nancial crisis, a large number of EMEs loosened monetary policy to cushion the e⁄ect of the global (cid:133)nancial crisis. This was a remarkable departure from previous crisis episodes during which EMEs had to tighten monetary policy. In this study, we explored the factors that enabled this shift in policy stance and (cid:133)nd statistically signi(cid:133)cant linkages between some characteristics of the economies and their ability to conduct countercyclical monetary policy. The results indicate that stronger macroeconomic fundamentals and reduced vulnerabilities, greater openness to trade and (cid:133)nancial (cid:135)ows, (cid:133)nancial reforms and the adoption of in(cid:135)ation targeting all facilitated the conduct of countercyclical policy in EMEs. Of these factors, (cid:133)nancial reforms and in(cid:135)ation targeting stand out as the most important. Several EMEs adopted in(cid:135)ation targeting since the late 1990s and, over the past decades, EMEs have also reformed their (cid:133)nancial sectors. In(cid:135)ation targeting regimes enhance transparency and (cid:135)exibility of monetary policy. By adopting in(cid:135)ation targeting and by implementing (cid:133)nancial reforms, EMEs also achieved a greater policy credibility. Indeed, lack of policy credibility is one of the main impediments to EMEs(cid:146)ability to conduct countercyclical monetary policy. When credibility is fragile, an attempt by the central bank to loosen monetary policy is perceived as a permanent switch to a loose money regime. This perception adversely a⁄ects con(cid:133)dence and increases risk premiums demanded by foreign investors. The adoption of in(cid:135)ation targeting helps to dispel these perceptions, as it fosters con(cid:133)denceinmonetarypolicyandanchorsin(cid:135)ationexpectations. Moreover,in(cid:135)ationtargetinghas been accompanied by reduced emphasis on exchange rate management, thereby allowing monetary policy to be (cid:135)exibly geared toward the stabilization of the domestic economy. We interpret our results to suggest that as long as EMEs maintain strong economic fundamentals, continue to reform their (cid:133)nancial markets, and adopt (cid:135)exible and transparent monetary policy frameworks such as in(cid:135)ation targeting, the conduct of countercyclical policy as an economic stabilization tool will likely be sustainable. The increasing popularity of in(cid:135)ation targeting among EMEs is particularly encouraging in this regard. We are not aware of a country that has adopted in(cid:135)ation targeting and abandoned it out of dissatisfaction, and there appears to be a degree of 15

irreversibility in (cid:133)nancial reforms, suggesting that the conduct of countercyclical policy could be sustainable. The increasing ability of EMEs to use monetary policy as a macroeconomic stabilization tool might partly explain the greater resilience of these economies to shocks emanating from the advanced economies despite increasing integration between the two sets of countries. 16

7 Appendix: Variables and Data Sources INTEREST RATES International Financial Statistics (IFS), Haver Analytics CENTRAL BANK CREDIT IFS INF IFS, World Development Indicators database (WDI) FXR2GDP IFS, WDI MATEXTDT WDI CAB2GDP WDI CGD2GDP WDI and IMF Historical Public Debt database STDT2EXTDT IFS, Global Development Finance database STDT2FXR Global Development Finance database, IFS OPENTRADE WDI OPENFIN Chinn-Ito Index database IT National sources EXCHREG Ilzetzki, Reinhart, and Rogo⁄(2008) database FINREF Abiad et. al (2010) Financial Reforms database FINDEV Constructed by authors using the following WDI and IFS data: bank deposit, (cid:133)nancial system deposit, deposit bank assets, private credit, and bank credit variables as percent of GDP credit data are obtained from WDI; bank data are obtained from IFS. FXR2GDP: Foreign exchange reserves to GDP ratio; CAB2GDP: Current account balance to GDP ratio; CGD2GDP: Central government debt to GDP ratio; STDT2EXTDT: Short-term debt to total debt ratio; STDT2FXR: Short-term debt to foreign exchange reserves ratio; EXTDT2EXP: External debt to GDP ratio; OPENTRADE: Trade volume to GDP ratio; MATEXTDT: Maturity of newly issued external debt in years; OPENFIN: Index for openness of the capital account; IT: Binary indicator for in(cid:135)ation targeting; EXCHREG: Exchange rate regime; FINREF: Index for (cid:133)nancial reforms; FINDEV: Index for (cid:133)nancial development. 17

References Abiad, A. and D. Enrica and T. Thierry (2008). "A New Database of Financial Reforms." IMF Working Paper No. 266. Aguiar,M.,andG.Gopinath(2007)."EmergingMarketBusinessCycles:TheCycleistheTrendm," Journal of Political Economy, Vol. 115 (1), pp. 69-102. Alesina, Alberto, and Lawrence H. Summers (1993). "Central Bank Independence and Macroeconomic Performance: Some Comparative Evidence," Journal of Money, Credit and Banking, Vol. 25 (May), pp. 151-62. Braun,Miguel(2001)."WhyisFiscalPolicyProcyclicalinDevelopingCountries?"mimeo,Harvard University. Calderon,C.andD.RobertoandS.Klaus(2003)."TheRoleofCredibilityintheCyclicalProperties of Macroeconomic Policies in Emerging Economies," Central Bank of Chile Working Paper No. 237. Callabero, Ricardo (2002). "Emerging Markets(cid:146)Economics", Mimeo, MIT. Calvo, Guillermo and Reinhart, Carmen (2002). "Fear of Floating," Quarterly Journal of Economics, Vol. 107, pp. 379-408. Chinn, Menzie and Hiro, Ito (2006). "A New Measure of Financial Openness," Journal of Comparative Policy Analysis 10(3). September 2008: pp. 307-320. Devereux, Michael and Lane, Philip (2003). "Understanding Bilateral Exchange Rate Volatility," CEPR Discussion Papers 3518. Flood, Robert and Jeanne, Olivier (2000). "An Interest Rate Defense of a Fixed Exchange Rate?" IMF Working Paper No. 159. Frankel, Je⁄rey and Rose, Andrew (1996). "Currency Crashes in emerging markets: An empirical treatment." Journal of International Economics, Vol. 41, pp. 351-366. Gavin, Michael and Perotti, Roberto (1997). "Fiscal Policy in Latin America." NBER Macroeconomics Annual 1997, Vol. 12, pp. 11-72. 18

Gupta, Sanjeev and Akitoby, Bernardin and Clements, Benedict and Inchauste Gabriela (2004). "The Cyclical and Long-Term Behavior of Government Expenditures in Developing Countries," IMF Working Paper No. 202. Hnatkovska, Viktoria and Lahiri Amartya and Vegh, Carlos (2008). "Interest Rates and the Exchange Rate: A Non-Montonic Tale," NBER Working Paper No. 13925. Ilzetzki, Ethan and Reinhart, Carmen and Rogo⁄, Kenneth (2011). "Exchange Rate Arrangements Entering the 21st Century: Which Anchor Will Hold?" University of Maryland Working Paper Series. Kaminsky, Graciela, Reinhart, Carmen and Vegh, Carlos (2004). "When it Rains, it Pours: Procyclical Capital Flows and Macroeconomic Policies," NBER Working Paper No. 10780. Lahiri A. and C. Vegh (200). "Delaying the Inevitable: Interest Rate Defense and Balance of Payments Crises," Journal of Political Economy 111. Lane, Philip (2003). "Business Cycles and Macroeconomic Policy in Emerging Market Economies," International Finance, 6 (1), pp. 89-108. Mendoza, Enrique (2002). "Why Should Emerging Economies Give Up National Currencies: A Case for Institutions Substitution," NBER Working Paper No. 8950. Milesi-Ferretti, Gian Maria and Razin, Assaf (2008). "Sharp Reductions in Current Account De(cid:133)cits: An Empirical Analysis," European Economic Review, Vol. 24, Issues 3-5, pp. 897-908. Mishkin, Frederic(2010)."MonetaryPolicyStrategy:LessonsFromtheCrisis,"inMonetaryPolicy Revisited: Lessons from the Crisis (European Central Bank: Frankfurt, forthcoming). Riascos, Alvaro and Vegh, Carlos (2003). "Procyclical Government Spending in Developing Countries: The Role of Capital Market Imperfections," UCLA Working Paper. Talvi, Ernesto and Vegh, Carlos (2004). "Tax Base Variability and Procyclical Fiscal Policy in Developing Countries," Journal of Development Economics, Vol. 78 pp. 156-190. Yakhin, Yossi (2008). "Financial Integration and Cyclicality of Monetary Policy in Small Open Economies," Monaster Center for Economic Research Discussion Paper No. 08-11. 19

02 51 01 5 0 Median Inflation (%) 1975 1980 1985 1990 1995 2000 2005 0 2 4 6 8 01 Median Current Account Balance (% of GDP) 1975 1980 1985 1990 1995 2000 2005 02 51 01 5 0 Median Foreign Exchange Reserves (% of GDP) 1975 1980 1985 1990 1995 2000 2005 non CCMP CCMP 08 06 04 02 Median Government Debt (% of GDP) 1975 1980 1985 1990 1995 2000 2005 non CCMP CCMP Figure 1: Timeline of Macroeconomic Variables by Monetary Policy Behavior during the 2008-2009 Crisis 20

51 01 5 0 Median Short term Debt (% of Total Debt) 1975 1980 1985 1990 1995 2000 2005 1 8. 6. 4. 2. 0 Median Short term debt (% of FX Reserves) 1975 1980 1985 1990 1995 2000 2005 1 5. 0 5. 1 Mean Capital Account Openness 1975 1980 1985 1990 1995 2000 2005 non CCMP CCMP 001 08 06 04 Median Trade Openness 1975 1980 1985 1990 1995 2000 2005 non CCMP CCMP Figure 2: Timeline of Macroeconomic Variables by Monetary Policy Behavior during the 2008-2009 Crisis 21

1 8. 6. 4. 2. Mean Currency peg 1975 1980 1985 1990 1995 2000 2005 2. 51. 1. 50. 0 Mean Inflation Targeting 1975 1980 1985 1990 1995 2000 2005 0 2. 4. 6. 8. Median Financial Development 1975 1980 1985 1990 1995 2000 2005 non CCMP CCMP 6. 5. 4. 3. 2. 1. Median Financial Reform 1975 1980 1985 1990 1995 2000 2005 non CCMP CCMP Figure 3: Timeline of Macroeconomic Variables by Monetary Policy Behavior during the 2008-2009 Crisis 22

sesirC fo rebmuN 08 06 04 02 0 1970 1975 1980 1985 1990 1995 2000 2005 2010 Emerging Markets Advanced Economies Figure 4: Frequency of Financial and Economic Crises: 1970-2009. yciloP yratenoM lacilcycretnuoC fo noitcarF 1 8. 6. 4. 2. 0 1970s 1980s 1990s 2000s 2008 2009 Advanced Emerging Economies Markets Figure 5: Fraction of Crises during which Advanced and Emerging Market Economies Conducted Countercyclical Monetary Policy. 23

selbairaV fo scitsitatS evitpircseD :1 elbaT 9002-0791 :elpmaS 9002-8002 :elpmaS elbairaV sbo .N .veD .tS naeM naideM sboN .veD .tS naeM naideM 771,1 18.366 36.921 41.11 061 46.976 24.06 15.5 FNI 1201 23.01 13.8 90.5 051 46.71 07.22 91.81 PDG2RXF 268 01.01 39.4- 11.4- 731 26.31 90.3- 04.4- PDG2BAC 529 38.87 80.17 86.15 451 62.64 44.84 09.73 PDG2DGC 359 11.41 63.31 87.9 721 38.41 16.51 03.21 TDTXE2TDTS 668 56.111 52.61 57.0 121 25.2 27.0 22.0 RXF2TDTS 711,1 11.24 18.37 05.56 051 76.35 53.99 39.98 EDARTNEPO 340,1 04.1 93.0- 31.1- 551 26.1 92.0 780.0- NIFNEPO 933,1 11.0 210.0 0 461 03.0 790.0 0 TI 179 66.1 56.2 2 141 59.0 58.1 2 GERHCXE 644 72.0 63.0 13.0 86 41.0 47.0 37.0 FERNIF 148 76.0 23.0- 05.0- 141 30.1 61.0 51.0- VEDNIF tnuoccatnerruC:PDG2BAC .oitar PDGotsevreseregnahcxengieroF:PDG2RXF .noita(cid:135)nI:FNI .raeysisirc-erpniderusaemeraselbairavllA .oitar tbed lanretxe latot ot tbed mret-trohS :TDTXE2TDTS .oitar PDG ot tbed tnemnrevog lartneC :PDG2DGC .oitar PDG ot ecnalab fossenneporofxednI:NIFNEPO .oitarPDGotemulovedarT:EDARTNEPO .oitarsevreseregnahcxengierofottbedmret-trohS:RXF2TDTS :FERNIF .)elbixe(cid:135) tsom=5 ,digir tsom=1( emiger etar egnahcxE :GERHCXE .gnitegrat noita(cid:135)ni rof rotacidni yraniB :TI .tnuocca latipac eht laicnan(cid:133)rofxednI:VEDNIF .)smroferlaicnan(cid:133)foleveltsehgih=1,smroferlaicnan(cid:133)foleveltsewol=0(sisircehtotpusmroferlaicnan(cid:133)rofxednI .snoitaluclac(cid:146)srohtua dna )xidneppA ees( secruos atad suoiraV :ecruoS .tnempoleved laicnan(cid:133) retaerg etacidni srebmun rehgiH .tnempoleved 24

noitcnuF yciloP yratenoM lacilcycretnuoC fo setamitsE tigoL :2 elbaT elbairaV sbo .N oitaR-sddO eulaV-P .feoC YTILIBARENLUV DNA SLATNEMMADNUF CIMONOCEORCAM 031 934.0 360.0 428.0- HGIH -FNI 621 62.2 650.0 718.0 WOL - PDG2DGC 811 64.3 600.0 142.1 HGIH - PDG2BAC 621 98.1 021.0 636.0 HGIH - PDG2RXF 201 06.0 703.0 015.0- WOL - TDTXE2TDTS 201 27.0 684.0 723.- HGIH -RXF2TDTS SSENNEPO 621 30.3 310.0 901.1 HGIH - NIFNEPO 221 54.2 240.0 698.0 HGIH - EDARTNEPO YTILIBIDERC DNA ;TNEMEGNARRA EGNAHCXE ;YRATENOM 921 38.1 503.0 506. TAOLF - GERHCXE 231 85.7 010.0 520.2 TI SMROFER DNA TNEMPOLEVED LAICNANIF 26 64.4 530.0 94.1 HGIH - FERNIF 221 67.1 561.0 665.0 HGIH - VEDNIF :raey sisirc-erp ni derusaem era selbairav tnednepedni llA .elbairav tnedneped eht si )PMCC( ycilop yratenom lacilcyc-retnuoc rof rotacidnI lartneC :PDG2DGC .oitar PDG ot ecnalab tnuocca tnerruC :PDG2BAC .oitar PDG ot sevreser egnahcxe ngieroF :PDG2RXF .noita(cid:135)nI :FNI egnahcxengierofottbedmret-trohS:RXF2TDTS .oitartbedlanretxelatotottbedmret-trohS:TDTXE2TDTS .oitarPDGottbedtnemnrevog rof rotacidni yraniB :TI .tnuocca latipac eht fo ssennepo rof xednI :NIFNEPO .oitar PDG ot emulov edarT :EDARTNEPO .oitar sevreser sisirc eht ot pu smrofer laicnan(cid:133) rof xednI :FERNIF .)elbixe(cid:135) tsom=5 ,digir tsom=1( emiger etar egnahcxE :GERHCXE .gnitegrat noita(cid:135)ni setacidni rebmun rehgiH .tnempoleved laicnan(cid:133) rof xednI :VEDNIF .)smrofer laicnan(cid:133) fo level tsehgih=1 ,smrofer laicnan(cid:133) fo level tsewol=0( .elitrauq poT :"HGIH" ;selitrauq elddiM :"DIM" ;elitrauq mottoB :"WOL" .tnempoleved laicnan(cid:133) retaerg 25

noitcnuF yciloP yratenoM lacilcycretnuoC fo setamitsE tigoL morf soitaR sddO :3 elbaT elbairaV )5( )4( )3( )2( )1( *75.2 *45.2 *85.2 **76.2 WOL - PDG2DGC 69.1 58.1 91.2 58.1 72.2 WOL - NIFNEPO **83.3 *42.3 **30.4 *71.3 **53.3 HGIH - NIFNEPO *12.0 *82.0 **52.0 WOL - EDARTNEPO 05.0 75.0 94.0 DIM - EDARTNEPO *45.2 04.2 HGIH - EDARTNEPO ***80.24 ***15.72 ***41.92 ***39.72 ***10.73 TI 08.0 WOL - VEDNIF 91.1 DIM - VEDNIF 90.1 93.1 HGIH - VEDNIF 311 311 811 311 811 N .raey sisirc-erp ni derusaem era selbairav tnednepedni llA .elbairav tnedneped eht si )PMCC( ycilop yratenom lacilcyc-retnuoc rof rotacidnI lartneC :PDG2DGC .oitar PDG ot ecnalab tnuocca tnerruC :PDG2BAC .oitar PDG ot sevreser egnahcxe ngieroF :PDG2RXF .noita(cid:135)nI :FNI egnahcxengierofottbedmret-trohS:RXF2TDTS .oitartbedlanretxelatotottbedmret-trohS:TDTXE2TDTS .oitarPDGottbedtnemnrevog rof rotacidni yraniB :TI .tnuocca latipac eht fo ssennepo rof xednI :NIFNEPO .oitar PDG ot emulov edarT :EDARTNEPO .oitar sevreser sisirc eht ot pu smrofer laicnan(cid:133) rof xednI :FERNIF .)elbixe(cid:135) tsom=5 ,digir tsom=1( emiger etar egnahcxE :GERHCXE .gnitegrat noita(cid:135)ni etacidni srebmun rehgiH .tnempoleved laicnan(cid:133) rof xednI :VEDNIF .)smrofer laicnan(cid:133) fo level tsehgih=1 ,smrofer laicnan(cid:133) fo level tsewol=0( .elitrauq poT :"HGIH" ;selitrauq elddiM :"DIM" ;elitrauq mottoB :"WOL" .tnempoleved laicnan(cid:133) retaerg 26

noticnuF yciloP yratenoM acilcycretnuoC fo setamitsE tigoL :4 elbaT elbairaV sboN oitaR-sddO eulaV-P .feoC YTILIBARENLUV DNA SLATNEMMADNUF CIMONOCEORCAM 648 26.1 300.0 184.0 WOL -FNI 286 41.1 944.0 531.0 HGIH - PDG2BAC 287 48.0 403.0 961.0- WOL - PDG2DGC 818 25.2 000.0 529.0 HGIH - PDG2RXF 227 88.0 384.0 031.0- WOL - TDTXE2TDTS 407 86.0 820.0 293.0- HGIH -RXF2TDTS SSENNEPO 097 54.1 110.0 373. HGIH - NIFNEPO 418 227.0 740.0 523.- WOL - EDARTNEPO YTILIBIDERC DNA ;TNEMEGNARRA EGNAHCXE ;YRATENOM 107 802.1 643.0 981.0 TAOLF - GERHCXE 278 628.6 210.0 129.1 TI SMROFER DNA TNEMPOLEVED LAICNANIF 987 502.2 000.0 197.0 HGIH - FERNIF 017 105.1 020.0 604.0 HGIH - VEDNIF .raey sisirc-erp ni derusaem era selbairav tnednepedni llA .elbairav tnedneped eht si )PMCC( ycilop yratenom lacilcyc-retnuoc rof rotacidnI lartneC :PDG2DGC .oitar PDG ot ecnalab tnuocca tnerruC :PDG2BAC .oitar PDG ot sevreser egnahcxe ngieroF :PDG2RXF .noita(cid:135)nI :FNI egnahcxengierofottbedmret-trohS:RXF2TDTS .oitartbedlanretxelatotottbedmret-trohS:TDTXE2TDTS .oitarPDGottbedtnemnrevog rof rotacidni yraniB :TI .tnuocca latipac eht fo ssennepo rof xednI :NIFNEPO .oitar PDG ot emulov edarT :EDARTNEPO .oitar sevreser sisirc eht ot pu smrofer laicnan(cid:133) rof xednI :FERNIF .)elbixe(cid:135) tsom=5 ,digir tsom=1( emiger etar egnahcxE :GERHCXE .gnitegrat noita(cid:135)ni etacidni srebmun rehgiH .tnempoleved laicnan(cid:133) rof xednI :VEDNIF .)smrofer laicnan(cid:133) fo level tsehgih=1 ,smrofer laicnan(cid:133) fo level tsewol=0( .elitrauq poT :"HGIH" ;selitrauq elddiM :"DIM" ;elitrauq mottoB :"WOL" .tnempoleved laicnan(cid:133) retaerg 27

noitcnuF yciloP yratenoM lacilcycretnuoC fo setamitsE tigoL morf soitaR sddO :5 elbaT elbairaV )21( )11( )01( )9( )8( )7( )6( )5( )4( )3( )2( )1( *51.2 97.1 *10.2 *41.2 *60.2 *79.1 ***63.2 ***01.2 ***43.2 ***33.2 ***03.2 ***92.2 WOL - FNI 91.1 91.1 31.1 02.1 22.1 10.1 10.1 89.0 80.1 89.0 HGIH - PDG2BAC 06.0 66.0 75.0 75.0 06.0 06.0 19.0 20.1 58.0 09.0 68.0 49.0 HGIH - PDG2DGC 26.1 **09.1 HGIH - PDG2RXF 39.0 30.1 WOL - TDTXE2TDTS 27.1 *81.2 *21.2 52.1 **38.1 **48.1 WOL - RXF2TDTS 94.0 *06. DIM - RXF2TDTS 47.0 34.0 **06. **24. HGIH - RXF2TDTS 01.1 21.1 81.1 72.1 21.1 38.0 27.0 78.0 68.0 19.0 DIM - NIFNEPO 61.1 *35.1 HGIH - NIFNEPO 90.1 40.1 22.1 31.1 89.0 71.1 24.1 23.1 14.1 02.1 71.1 82.1 HGIH - EDARTNEPO *36.5 *56.5 *21.6 *27.5 93.5 *72.6 **94.9 **90.01 **97.01 **84.8 *26.8 **00.11 TI 22.1 62.1 61.1 51.1 81.1 71.1 00.1 29.0 09.0 49.0 49.0 49.0 HGIH - GERHCXE 79.0 39.0 49.0 10.1 39.0 94.1 14.1 *65.1 *46.1 *06.1 WOL - VEDNIF 87.0 *86. DIM - VEDNIF ***20.3 ***17.3 ***21.3 ***89.2 ***97.2 ***39.2 HGIH - FERNIF 142 852 142 142 142 142 404 764 404 404 304 404 N .raey sisirc-erp ni derusaem era selbairav tnednepedni llA .elbairav tnedneped eht si )PMCC( ycilop yratenom lacilcyc-retnuoc rof rotacidnI lartneC :PDG2DGC .oitar PDG ot ecnalab tnuocca tnerruC :PDG2BAC .oitar PDG ot sevreser egnahcxe ngieroF :PDG2RXF .noita(cid:135)nI :FNI egnahcxengierofottbedmret-trohS:RXF2TDTS .oitartbedlanretxelatotottbedmret-trohS:TDTXE2TDTS .oitarPDGottbedtnemnrevog rof rotacidni yraniB :TI .tnuocca latipac eht fo ssennepo rof xednI :NIFNEPO .oitar PDG ot emulov edarT :EDARTNEPO .oitar sevreser sisirc eht ot pu smrofer laicnan(cid:133) rof xednI :FERNIF .)elbixe(cid:135) tsom=5 ,digir tsom=1( emiger etar egnahcxE :GERHCXE .gnitegrat noita(cid:135)ni setacidni rebmun rehgiH .tnempoleved laicnan(cid:133) rof xednI :VEDNIF .)smrofer laicnan(cid:133) fo level tsehgih=1 ,smrofer laicnan(cid:133) fo level tsewol=0( .elitrauq poT :"HGIH" ;selitrauq elddiM :"DIM" ;elitrauq mottoB :"WOL" .tnempoleved laicnan(cid:133) retaerg 28

noitcnuF yciloP yratenoM lacilcycretnuoC fo setamitsE tigoL morf soitaR sddO :sisylanA ssentsuboR :6 elbaT elbairaV )4( )3( )2( )1( 74.1 *95.1 92.1 17.0 WOL - FNI DIM - FNI **20.2 *37.1 **69.1 *33.2 HGIH - PDG2RXF *65.0 38.0 53.1 15.1 WOL - PDG2DGC DIM - PDG2DGC HGIH - PDG2DGC 11.1 14.1 17.0 *67.2 WOL - RXF2TDTS DIM - RXF2TDTS WOL - EDARTNEPO DIM - NIFNEPO 74.1 39.0 49.0 HGIH -NIFNEPO **51.01 *47.8 **44.9 *30.11 TI DIM - FERNIF *97.1 *98.1 ***39.5 *86.2 HGIH - FERNIF 13.1 **88.1 ***73.2 **54.2 WOL -VEDNIF DIM - VEDNIF 75.0 *55.0 57.0 TAOLF 203 204 183 971 N .raey sisirc-erp ni derusaem era selbairav tnednepedni llA .elbairav tnedneped eht si )PMCC( ycilop yratenom lacilcyc-retnuoc rof rotacidnI lartneC :PDG2DGC .oitar PDG ot ecnalab tnuocca tnerruC :PDG2BAC .oitar PDG ot sevreser egnahcxe ngieroF :PDG2RXF .noita(cid:135)nI :FNI egnahcxengierofottbedmret-trohS:RXF2TDTS .oitartbedlanretxelatotottbedmret-trohS:TDTXE2TDTS .oitarPDGottbedtnemnrevog rof rotacidni yraniB :TI .tnuocca latipac eht fo ssennepo rof xednI :NIFNEPO .oitar PDG ot emulov edarT :EDARTNEPO .oitar sevreser sisirc eht ot pu smrofer laicnan(cid:133) rof xednI :FERNIF .)elbixe(cid:135) tsom=5 ,digir tsom=1( emiger etar egnahcxE :GERHCXE .gnitegrat noita(cid:135)ni etacidni srebmun rehgiH .tnempoleved laicnan(cid:133) rof xednI :VEDNIF .)smrofer laicnan(cid:133) fo level tsehgih=1 ,smrofer laicnan(cid:133) fo level tsewol=0( otelpmasehttcirtserew :1nmuloC .elitrauqpoT :"HGIH";selitrauqelddiM :"DIM";elitrauqmottoB :"WOL" .tnempolevedlaicnan(cid:133)retaerg evomer ew :3 nmuloC .ylno setar ycilop ot setar tseretni mret-trohs fo noitin(cid:133)ed eht tcirtser ew :2 nmuloC .semiger etar egnahcxe deggep-non .noitavresbo dnoces eht pord ew ,yrtnuoc nevig a rof sraey evitucesnoc owt ni srucco sisirc a nehw :4 nmuloC .ycilop "lacilcyca" htiw seirtnuoc 29